Live broadcast manuscript generation method and device, storage medium and program product

Abstract

The embodiment of the invention provides a live broadcast manuscript generation method and device, a storage medium and a program product. In the method, promotion demand information of a merchant can be obtained, and the promotion demand information comprises basic information of the merchant, target object description information and a manuscript style; inputting the model cue word and the promotion demand information into a manuscript generation model, generating an initial live broadcast manuscript according with the manuscript style according to the basic information and the description information, and generating a plurality of live broadcast verbal skills required by the promotion target object under the guidance of a plurality of verbal skill cue words, adding various live broadcast verbal skills into the initial live broadcast manuscript to obtain an intermediate-state live broadcast manuscript; and performing regularization modification on the intermediate-state live file to obtain a target live file. In this way, the initial live broadcast manuscript conforming to the manuscript style can be accurately generated by using the manuscript generation model, and then verbal skill enhancement can be performed under the guidance of various verbal skill cues so as to accurately obtain the final live broadcast manuscript.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, storage medium and program product for generating a live broadcast script. Background Art

[0002] With the rapid development of e-commerce, merchants have an increasing demand for the promotion of stores or products. With the development of digital human technology, many e-commerce platforms have begun to use digital human live broadcast as an important means of promotion. The promotion method based on digital humans requires the live broadcast script to drive the digital humans, thereby achieving the purpose of promoting and introducing merchants' stores or products through digital humans. In the existing technology, the generation of live broadcast scripts mainly relies on manual design, with low generation efficiency, and it is difficult to meet the growing promotion needs. Summary of the invention

[0003] Multiple aspects of the present application provide a live broadcast script generation method, device, storage medium and program product for generating a live broadcast script more accurately.

[0004] An embodiment of the present application provides a method for generating a live broadcast script, including: obtaining promotion demand information of a merchant, the promotion demand information including basic information of the merchant, description information of a target object, and a target manuscript style selected by the merchant from a plurality of manuscript styles, the target object being an object to be promoted by the merchant; inputting model prompt words and the promotion demand information into a manuscript generation model based on artificial intelligence AI, generating an initial live broadcast script that conforms to the target manuscript style according to the basic information of the merchant and the description information of the target object, and generating a plurality of live broadcast scripts required for promoting the target object under the guidance of a plurality of script prompt words included in the model prompt words, adding the plurality of live broadcast scripts to a plurality of end symbol positions in the initial live broadcast script to obtain an intermediate live broadcast script; and modifying the intermediate live broadcast script in a regularized manner based on the preset scripts of the merchant to obtain a target live broadcast script.

[0005] Further optionally, obtaining the merchant's promotion demand information includes: responding to the merchant's live broadcast script generation instruction, displaying a script service page for the merchant to determine the target object to be promoted; based on the target type of the target object, displaying multiple information items corresponding to the target type on the script service page for the merchant to input information; responding to the merchant's information input operation, obtaining the merchant's basic information, the target object's description information, and the target script style selected by the merchant from a plurality of script styles.

[0006] Further optionally, the multiple speech prompts include a first category of speech prompts and a second category of speech prompts, and each category of speech prompts includes at least one speech prompt; the manuscript generation model includes: a manuscript generation network and a speech enhancement network; the model prompts and the promotion demand information are input into the manuscript generation model based on artificial intelligence AI, and an initial live manuscript that conforms to the target manuscript style is generated according to the basic information of the merchant and the description information of the target object, and under the guidance of the multiple speech prompts, a variety of live scripts required for promoting the target object are generated, and the multiple live scripts are added to multiple end mark positions in the initial live manuscript to obtain an intermediate live manuscript, including: adding the merchant's The basic information and the descriptive information of the target object are input into the script generation network, and an initial live script that conforms to the style of the target script is generated according to the basic information of the merchant and the descriptive information of the target object; the multiple speech prompts and the basic information of the merchant are input into the speech enhancement network, and under the guidance of the first type of speech prompts, at least one first live script is generated based on the basic information of the merchant; and under the guidance of the second type of speech prompts, at least one second live script is generated based on the merchant's custom knowledge base; the at least one first live script and the at least one second live script are added to multiple end symbol positions in the initial live script to obtain the intermediate live script.

[0007] Further optionally, the model prompt words also include key information prompt words; the basic information of the merchant and the descriptive information of the target object are input into the manuscript generation network, and an initial live manuscript that conforms to the target manuscript style is generated according to the basic information of the merchant and the descriptive information of the target object, including: extracting features from the basic information of the merchant and the descriptive information of the target object to obtain promotion demand features; the promotion demand features include: merchant features corresponding to the basic information and object features corresponding to the descriptive information of the target object; under the guidance of the key information prompt words, the promotion demand features are divided into key demand features and non-key demand features; according to the key demand features and the non-key demand features, an initial live manuscript that conforms to the target manuscript style is generated; the initial live manuscript focuses on the manuscript segment obtained based on the mapping of the key demand features.

[0008] Further optionally, each speech prompt word includes multiple speech examples and speech control information, and the speech control information is at least used to limit the number of words, style and type of the speech; under the guidance of the first category of speech prompt words, at least one first live broadcast speech is generated based on the basic information of the merchant, including: for each speech prompt word in the first category of speech prompt words, the speech generation capability of the corresponding style is learned from multiple speech examples in the speech prompt word, and the first live broadcast speech is generated according to the requirements of the speech control information in the speech prompt word by using the learned speech generation capability of the corresponding style.

[0009] Further optionally, the merchant's custom knowledge base has pre-learned the merchant's language habits based on the merchant's historical speech, the merchant's detailed information and the target object's detailed information; each speech prompt word includes multiple speech examples and speech control information, and the speech control information is at least used to limit the number of words, style and type of the speech; under the guidance of the second category of speech prompt words, at least one second live speech is generated based on the merchant's custom knowledge base, including: for each speech prompt word in the second category of speech prompt words, learning the speech generation capability of the corresponding style from multiple speech examples in the speech prompt word; based on the merchant's language habits learned from the custom knowledge base, generating at least one basic speech fragment matching the merchant; using the learned speech generation capability of the corresponding style, generating the at least one second live speech according to the requirements of the speech control information in the speech prompt word and the at least one basic speech fragment.

[0010] Further optionally, the multiple live scripts are added to multiple end-marker positions in the initial live script to obtain an intermediate live script, including: traversing the initial live script to determine the position of the end mark in the initial live script, the end mark includes punctuation marks used to describe the end of a sentence or a paragraph; randomly adding the multiple live scripts to multiple end-marker positions in the initial live script to obtain the intermediate live script.

[0011] Further optionally, it also includes: obtaining sample promotion demand information and a sample live broadcast script with added sample live broadcast script; inputting the sample promotion demand information and the sample live broadcast script into a preset script generation network, under the supervision of the sample live broadcast script, with the loss function of the script generation network converging to a target range as the goal, using the sample promotion demand information to train the script generation network to obtain a script generation model; wherein the loss function is used to calculate the error between the live broadcast script generated by the script generation network based on the sample promotion demand information and the sample live broadcast script.

[0012] An embodiment of the present application also provides an electronic device, comprising: a memory and a processor; wherein the memory is used to: store one or more computer instructions; the processor is used to execute the one or more computer instructions to: execute the steps in the live script generation method.

[0013] An embodiment of the present application also provides a computer-readable storage medium, which, when the computer program is executed by a processor, enables the processor to implement the steps in the live broadcast script generation method.

[0014] An embodiment of the present application also provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the processor is enabled to implement the steps in the live script generation method.

[0015] In this embodiment, the merchant's promotion demand information can be obtained, and the promotion demand information includes the merchant's basic information, target object description information, and manuscript style; the model prompt words and the promotion demand information are input into the manuscript generation model, and an initial live manuscript that conforms to the manuscript style is generated according to the basic information and description information, and under the guidance of a variety of script prompt words, a variety of live scripts required for the promotion target object are generated, and a variety of live scripts are added to the initial live manuscript to obtain an intermediate live manuscript; the intermediate live manuscript is modified in a regular manner to obtain a target live manuscript. In this way, the manuscript generation model can be used to more accurately generate an initial live manuscript that conforms to the manuscript style, and then the script can be enhanced under the guidance of a variety of script prompt words to more accurately obtain the final live manuscript. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A flowchart of a method for generating a live broadcast script provided by an exemplary embodiment of the present application;

[0018] Figure 2 A flowchart of a method for generating a live broadcast script in an actual scenario provided by an exemplary embodiment of the present application;

[0019] Figure 3 A schematic diagram of an electronic device provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0021] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation portals for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) are in compliance with relevant laws and standards.

[0022] With the rapid development of e-commerce, merchants have an increasing demand for the promotion of stores or products. With the development of digital human technology, many e-commerce platforms have begun to use digital human live broadcast as an important means of promotion. The promotion method based on digital humans requires the live broadcast script to drive the digital humans, thereby achieving the purpose of promoting and introducing merchants' stores or products through digital humans. In the existing technology, the generation of live broadcast scripts mainly relies on manual design, with low generation efficiency, and it is difficult to meet the growing promotion needs.

[0023] In view of the above technical problems, a technical solution is provided in the embodiments of the present application for generating live broadcast scripts for merchants more efficiently. The technical solutions provided in the embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0024] Figure 1 The live broadcast script generation method provided by the exemplary embodiment of the present application is as follows: Figure 1 As shown, the method may include the following steps:

[0025] Step 11: Obtain the merchant's promotion demand information, which includes the merchant's basic information, target object description information, and the target manuscript style selected by the merchant from a variety of manuscript styles. The target object is the object to be promoted by the merchant.

[0026] Step 12. Input the model prompt words and promotion demand information into the manuscript generation model based on artificial intelligence AI, generate an initial live manuscript that conforms to the target manuscript style according to the basic information of the merchant and the description information of the target object, and under the guidance of the various script prompt words included in the model prompt words, generate a variety of live scripts required to promote the target object, and add the various live scripts to multiple end symbol positions in the initial live manuscript to obtain an intermediate live manuscript.

[0027] Step 13: Based on the merchant's preset words, the intermediate live broadcast script is modified in a regular manner to obtain the target live broadcast script.

[0028] The execution subject in this embodiment can be any electronic device, such as a server or a terminal device. The terminal device can be a mobile phone, a tablet computer, a computer, etc. This embodiment does not limit the specific implementation method of the electronic device.

[0029] In this embodiment, the promotion demand information of the merchant may include the merchant's basic information, the description information of the target object, and the target manuscript style selected by the merchant from a plurality of manuscript styles. The merchant's basic information includes but is not limited to at least one of the following: the merchant's name and the merchant's introduction, etc. The target object is the object to be promoted by the merchant, which may be a service, such as housekeeping service and laundry service, etc., or a commodity, such as a table, a house, and a chair, etc. The description information of the target object is used to describe the details of the target object, including but not limited to at least one of the following: identification information, name information, brief introduction information, service item information, purchase rule information, etc.

[0030] In this embodiment, a variety of manuscript styles can be provided to merchants for selection. The specific implementation method of the manuscript style is not limited here, and any style can be set according to actual design requirements, such as lively and generous style, humorous style, friendly and natural style, professional and authoritative style, and urgent style.

[0031] In this embodiment, the model prompt words and promotion demand information can be input into the manuscript generation model based on artificial intelligence (AI). The manuscript generation model can first generate an initial live broadcast manuscript that meets the target manuscript style based on the basic information of the merchant and the description information of the target object.

[0032] Afterwards, the manuscript generation model can generate a variety of live broadcast scripts required for promotion to the target object under the guidance of the various script prompts included in the model prompts, and add the various live broadcast scripts to multiple end mark positions in the initial live broadcast manuscript to obtain an intermediate live broadcast manuscript.

[0033] Each speech prompt is used to guide the live speech generation of the manuscript generation model to reduce the deviation between the generated live speech and the ideal effect. In other words, the speech prompt can guide the manuscript generation model to generate live speech that better matches the promotion needs of the merchant. Each speech prompt may include at least one speech example; different types of speech prompts have different speech styles and different speech examples. Each speech prompt can instruct the manuscript generation model to generate speech in the corresponding speech style based on at least one speech example.

[0034] For example, the various speech prompts may include: welcome speech prompts, self-question and self-answer speech prompts, and purchase recommendation speech prompts. Among them, the welcome speech prompt may include two speech examples, namely "Welcome! We have carefully selected today's popular products for you, let's see if there are any products that you are interested in" and "Hello, based on your recent preferences, we found that these products may just suit your taste, why not come and learn more about them."

[0035] The above is only an exemplary description, and the various speech prompts may also include any type of speech prompts, which is not limited in the present embodiment; for example, speech prompts of the type of asking for needs and prompts of the type of closing the live broadcast; the prompts of the type of closing the live broadcast may include the following three speech examples, namely "Dear friends, our live broadcast today is about to end! If you still have your favorite products that you have not ordered, please hurry up! This opportunity is rare, if you miss it today, you may not get such a favorable price", "Thank you all for your enthusiastic participation and support! In the last few minutes, we have prepared a special gift for all friends who have not yet placed an order. Order now to enjoy extra discounts" and "Attention, friends, our flash sale is about to end! There are less than 10 minutes left until the end of the live broadcast, please act now and seize this last chance to enjoy super discounts."

[0036] The various live broadcast scripts generated in this way are adapted to the live broadcast scenarios, which can enrich the content of the manuscript, attract users and improve the readability of the manuscript.

[0037] Afterwards, the script generation model can add various live broadcast scripts to multiple end-of-sentence positions in the initial live broadcast script to obtain an intermediate live broadcast script. The end-of-sentence positions may include: the position of a period, the position of a greeting, and the position of an exclamation mark, etc., which are not limited in this embodiment. In other words, various live broadcast scripts can be inserted into the middle, end, and end of a sentence in the initial live broadcast script, etc., which are not limited in this embodiment.

[0038] Based on the above steps, the intermediate live broadcast manuscript is modified in a regular manner based on the merchant's preset words to obtain the target live broadcast manuscript. Among them, regular modification refers to modifying the intermediate live broadcast manuscript into a target live broadcast manuscript that conforms to the live broadcast scene rules. For example, the merchant's preset words are added to the intermediate live broadcast manuscript. The merchant's preset words are prompts that need to be added to the manuscript, such as delivery time. For another example, the format of the intermediate live broadcast manuscript is controlled, including: word count control, deletion of banned words, replacement of sensitive words, etc.

[0039] In this way, the script generation model can be used to more accurately generate an initial live script that conforms to the script style, and then the script can be enhanced under the guidance of a variety of script prompts to generate a variety of live scripts that are suitable for the live broadcast scene, thereby enriching the content of the initial live script, attracting users and improving the readability of the initial live script. Then, based on rule-based modifications, the final target live script can be obtained more accurately.

[0040] In some optional embodiments, when obtaining the merchant's promotion demand information, the merchant's live broadcast manuscript generation instruction can be responded to, and the manuscript service page can be displayed to the merchant for the merchant to determine the target object to be promoted. Among them, the manuscript service page can display multiple objects to be selected by the merchant. Afterwards, according to the target type of the target object, multiple information items corresponding to the target type can be displayed on the manuscript service page for the merchant to input information, wherein the type of the object can be arbitrarily divided according to the actual design requirements, and this embodiment is not limited. For example, it can be a store type, a product type, a service type, and so on. Among them, the type of the object is different, and the information items displayed on the manuscript service page are also different. Each type can have a corresponding relationship with the information item. Based on this, according to the corresponding relationship and the target type of the target object, multiple information items corresponding to the target type can be displayed on the manuscript service page for the merchant to input information.

[0041] Afterwards, the merchant's information input operation can be responded to to obtain the merchant's basic information, the target object's description information, and the target document style selected by the merchant from multiple document styles. The information input operation can be implemented as any type of operation. For example, the information input operation can be divided into two operations. One is an input operation, in which the user can perform an input operation to input the merchant's basic information and the target object's description information; the other is a drop-down list operation, in which the user can click the drop-down list and select the target document style from multiple document styles.

[0042] In this way, the promotion demand information of merchants can be obtained more accurately.

[0043] In some optional embodiments, the multiple speech prompt words may include a first type of speech prompt words and a second type of speech prompt words, and each type of speech prompt words may include at least one speech prompt word. The use of "first type" and "second type" to define speech prompt words is only used to distinguish speech prompt words with different functions. The manuscript generation model includes: a manuscript generation network and a speech enhancement network;

[0044] Based on this, step 12 of the aforementioned embodiment, "inputting the model prompt words and promotion demand information into the manuscript generation model based on artificial intelligence AI, generating an initial live manuscript that conforms to the target manuscript style according to the basic information of the merchant and the description information of the target object, and under the guidance of multiple script prompt words, generating multiple live scripts required for promoting the target object, and adding multiple live scripts to multiple end mark positions in the initial live manuscript to obtain an intermediate live manuscript", can be implemented based on the following steps:

[0045] Step 121: input the merchant's basic information and the target object's description information into the document generation network, and generate an initial live broadcast document that conforms to the target document style based on the merchant's basic information and the target object's description information.

[0046] In this step, the model prompt words also include key information prompt words, which are used to guide the manuscript generation network to select more important features from the more complex promotion demand features, namely, key demand features. Taking the live broadcast sales of product A as an example, among the more complex promotion demand features, the more important features may be the features of the selling points of product A. For example, if the selling point of product A is high cost performance, then the price in the description information of the target object is more important.

[0047] Based on this, in this step, feature extraction can be performed on the merchant's basic information and the target object's description information to obtain promotion demand features; the promotion demand features may include: merchant features corresponding to the basic information and object features corresponding to the target object's description information. Under the guidance of key information prompt words, the promotion demand features can be divided into key demand features and non-key demand features. Based on the key demand features and non-key demand features, an initial live broadcast manuscript that meets the target manuscript style is generated; the initial live broadcast manuscript highlights the manuscript segments obtained based on the mapping of key demand features. The emphasis here can be understood as: the manuscript segments obtained based on the mapping of key demand features account for a large proportion of the initial live broadcast manuscript, or the length is large, or the language style of the manuscript segments obtained based on the mapping of key demand features is more radical than the language style of the manuscript segments obtained based on the mapping of non-key demand features.

[0048] In this way, an initial live broadcast script that conforms to the target script style can be generated more accurately based on the merchant's basic information and the target object's descriptive information. The initial live broadcast script is more attractive for the audience to pay attention to the key information in the script.

[0049] Step 122: input a plurality of speech prompts and basic information of the merchant into the speech enhancement network, and generate at least one first live broadcast speech based on the basic information of the merchant under the guidance of the first type of speech prompts.

[0050] Each speech prompt word may include multiple speech examples and speech control information, and the speech control information may be used to limit the number of words, style and type of the speech. For example, the speech control information L1 may limit the speech to a maximum of 20 words, a humorous style and a welcome type. Based on the speech control information L1, the following first live broadcast speech with a humorous style and a welcome type may be obtained: "Welcome to the shopping paradise! Here, the only temptation you need to resist is - don't buy all the goods on the entire website home."

[0051] Afterwards, for each type of speech prompt words in the first type of speech prompt words, the speech generation capability of the corresponding style can be learned from multiple speech examples in the speech prompt words, and the first live broadcast speech can be generated according to the requirements of the speech control information in the speech prompt words by using the learned speech generation capability of the corresponding style. In this way, the first live broadcast speech can be generated more accurately under the demonstration of the speech examples and the constraints of the speech control information.

[0052] Step 123, under the guidance of the second type of speech prompt words, generate at least one second live speech based on the merchant's custom knowledge base. Among them, the merchant's custom knowledge base has already learned the merchant's language habits based on the merchant's historical speech, the merchant's detailed information and the detailed information of the target object. The detailed information here is different from the "basic information" and "description information" in the previous text, and can be used to describe more comprehensive information about the merchant and the target object. The merchant's historical speech can be obtained by extracting the merchant's historical live video or historical live audio; based on this, the merchant's custom knowledge base can be established by learning the merchant's historical speech, the merchant's detailed information and the detailed information of the target object, and the merchant's language habits can be learned. The merchant's language habits are also more consistent with the merchant and the target object's own detailed information.

[0053] Based on this, for each type of speech prompt in the second type of speech prompt, the speech generation capability of the corresponding style can be learned from multiple speech examples in the speech prompt; based on the language habits of the merchant learned in the custom knowledge base, at least one basic speech segment that matches the merchant is generated. In other words, the basic speech segment can be understood as a segment that imitates the tone and voice of the merchant and is generated based on the detailed information of the merchant and the target object.

[0054] Afterwards, the learned speech generation capability of the corresponding style can be used to generate at least one second live speech in accordance with the requirements of the speech control information in the speech prompt and at least one basic speech segment. That is to say, the learned speech generation capability of the corresponding style can be used to perform corresponding processing on the basis of at least one basic speech segment in accordance with the requirements of the speech control information in the speech prompt to obtain at least one second live speech. The corresponding processing here can be: expanding or modifying the basic speech segment, or generating a new basic speech segment using the learned speech generation capability of the corresponding style and connecting two adjacent basic speech segments using the new basic speech segment, which is not limited in this embodiment.

[0055] Step 124: Add at least one first live broadcast script and at least one second live broadcast script to multiple end mark positions in the initial live broadcast script to obtain an intermediate live broadcast script.

[0056] Among them, each live broadcast script can be added to the end mark position in the initial live broadcast script every time it is generated, or all live broadcast scripts can be added to the end mark position in the initial live broadcast script after they are generated. This embodiment does not limit this.

[0057] Based on the above steps 121 to 124, an intermediate live broadcast script can be generated more accurately based on the script generation model and the corresponding prompt words.

[0058] In some optional embodiments, when adding multiple live broadcast scripts to multiple end-of-speech positions in the initial live broadcast script to obtain an intermediate live broadcast script, the initial live broadcast script may be traversed to determine the position of the end-of-speech in the initial live broadcast script, where the end-of-speech includes punctuation marks used to describe the end of a sentence or paragraph, such as a period, a question mark, or an exclamation mark. Afterwards, the multiple live broadcast scripts may be randomly added to multiple end-of-speech positions in the initial live broadcast script to more accurately obtain an intermediate live broadcast script.

[0059] In some optional embodiments, the manuscript generation model in the aforementioned embodiments can be trained based on the following method: obtaining sample promotion demand information and a sample live broadcast manuscript with sample live broadcast script added; inputting the sample promotion demand information and the sample live broadcast manuscript into a preset manuscript generation network, and under the supervision of the sample live broadcast manuscript, with the loss function of the manuscript generation network converging to a target range as the goal, using the sample promotion demand information to train the manuscript generation network to obtain a manuscript generation model; wherein the loss function is used to calculate the error between the live broadcast manuscript generated by the manuscript generation network based on the sample promotion demand information and the sample live broadcast manuscript.

[0060] The following will further illustrate the live broadcast script generation method in the aforementioned embodiment in combination with actual scenarios.

[0061] like Figure 2 As shown, the embodiment of the present application can combine a large language model (i.e., a manuscript generation model based on artificial intelligence AI), a merchant-defined knowledge base, and a merchant-preset speech, etc., to more accurately output the manuscript, that is, generate a target live broadcast manuscript.

[0062] According to the different input promotion demand information, the script generation model can be used for intelligent analysis to generate live scripts suitable for different scenarios. The introduction of multi-level customization options ensures applicability and diversity, covering the needs of different types of merchants. Specifically, it may include the following:

[0063] (1) Obtaining the merchant's promotion demand information:

[0064] The generation of live broadcast scripts for products depends on the specific information of the products, including product identification number, product name, product introduction, service information, purchase rules and style reference information (i.e. target script style). Based on this, the script generation model can fully integrate this information in the script generation process to ensure that the audience is guided to pay attention to and purchase specific products. The generation of live broadcast scripts for stores depends on the store name, store introduction and style reference information provided by the merchant, and the live broadcast script can be automatically created based on the above information collected.

[0065] (2) Use the manuscript generation model to obtain the intermediate live manuscript:

[0066] ① Style definition: In order to enhance the diversity of manuscript styles, merchants can choose different manuscript styles according to their needs in the background, including but not limited to lively and generous, humorous and interesting, friendly and natural, professional and authoritative, and urgent styles. For example, urgency means limited inventory for some merchants. After the style is defined, the manuscript generation model is used to generate the initial live broadcast manuscript based on the promotion demand information.

[0067] ② Speech enhancement: In order to make the generated manuscript closer to the live broadcast environment and enrich the manuscript style, the readability and attractiveness of the initial live broadcast manuscript are improved by integrating various speech techniques with actual needs. This speech enhancement part can include the following two parts:

[0068] Part 1: Generate a variety of live broadcast scripts based on the basic information of the merchant through the script generation model. For example, it can include welcome-type live broadcast scripts, self-questioning and self-answering-type live broadcast scripts, and recommended purchase-type live broadcast scripts. A variety of live broadcast scripts are randomly inserted into the initial live broadcast script, and the insertion position can be at the beginning or end of a paragraph. The judgment criteria are Chinese paragraph end symbols, including periods, greetings, and exclamation marks. For example, the self-questioning and self-answering type of live broadcast script generates a self-questioning and self-answering similar interactive dialogue based on the details of the merchant's goods and services. When each script is generated, 3 examples will be provided to prevent generation deviation. Control the language conciseness in the prompt words to ensure that the audience grasps the product information in the shortest time.

[0069] Part 2: Use the manuscript generation model and combine it with the merchant’s customized knowledge base to generate three fluent and natural sentences from different angles, and randomly insert them into the initial live broadcast manuscript.

[0070] (3) Regularization modification: This part is the last step of adjustment before the document is output. The target live document can be obtained by regularizing the intermediate live document.

[0071] Based on the above content, the script generation model can be used to more accurately generate an initial live script that conforms to the script style, and then the speech enhancement can be performed under the guidance of a variety of speech prompts to more accurately obtain the final live script.

[0072] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 11 to 13 can be device A; for another example, the execution subject of steps 11 to 12 can be device A, and the execution subject of step 13 can be device B; and so on.

[0073] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel, and the sequence numbers of the operations, such as 12, 13, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel.

[0074] It should be noted that the descriptions such as “first” and “second” in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit “first” and “second” to different types.

[0075] Figure 3is a schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present application, and the electronic device is applicable to the live broadcast script generation method provided by the aforementioned embodiment, such as Figure 3 As shown, the electronic device may include: a memory 301 , a processor 302 and a communication component 303 .

[0076] The memory 301 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, contact data, phone book data, messages, pictures, videos, etc.

[0077] In some exemplary embodiments, the processor 302 is coupled to the memory 301 and is used to execute a computer program in the memory 301, so as to: obtain the promotion demand information of the merchant, the promotion demand information including the basic information of the merchant, the description information of the target object and the target manuscript style selected by the merchant from a plurality of manuscript styles, and the target object is the object to be promoted by the merchant; input the model prompt words and the promotion demand information into a manuscript generation model based on artificial intelligence AI, generate an initial live broadcast manuscript conforming to the target manuscript style according to the basic information of the merchant and the description information of the target object, and generate a plurality of live broadcast scripts required for promoting the target object under the guidance of a plurality of script prompt words included in the model prompt words, add the plurality of live broadcast scripts to a plurality of end symbol positions in the initial live broadcast manuscript to obtain an intermediate live broadcast manuscript; and modify the intermediate live broadcast manuscript in a regularized manner based on the preset script of the merchant to obtain a target live broadcast manuscript.

[0078] Further optionally, when the processor 302 obtains the merchant's promotion demand information, it is specifically used to: respond to the merchant's live broadcast script generation instruction, and display a script service page for the merchant to determine the target object to be promoted; based on the target type of the target object, display multiple information items corresponding to the target type on the script service page for the merchant to input information; respond to the merchant's information input operation, and obtain the merchant's basic information, the target object's description information, and the target script style selected by the merchant from a plurality of script styles.

[0079] Further optionally, the multiple speech prompts include a first category of speech prompts and a second category of speech prompts, and each category of speech prompts includes at least one speech prompt; the manuscript generation model includes: a manuscript generation network and a speech enhancement network; the processor 302 inputs the model prompts and the promotion demand information into the manuscript generation model based on artificial intelligence AI, generates an initial live manuscript that conforms to the target manuscript style according to the basic information of the merchant and the description information of the target object, and generates a variety of live broadcast scripts required to promote the target object under the guidance of the multiple speech prompts, and adds the multiple live broadcast scripts to multiple end mark positions in the initial live manuscript to obtain an intermediate live manuscript, which is specifically used to: The basic information of the merchant and the descriptive information of the target object are input into the manuscript generation network, and an initial live manuscript that conforms to the style of the target manuscript is generated according to the basic information of the merchant and the descriptive information of the target object; the multiple speech prompts and the basic information of the merchant are input into the speech enhancement network, and under the guidance of the first type of speech prompts, at least one first live speech is generated based on the basic information of the merchant; and under the guidance of the second type of speech prompts, at least one second live speech is generated based on the merchant's custom knowledge base; the at least one first live speech and the at least one second live speech are added to multiple end symbol positions in the initial live manuscript to obtain the intermediate live manuscript.

[0080] Further optionally, the model prompt words also include key information prompt words; the processor 302 inputs the basic information of the merchant and the descriptive information of the target object into the manuscript generation network, and generates an initial live manuscript that conforms to the target manuscript style based on the basic information of the merchant and the descriptive information of the target object, specifically for: extracting features from the basic information of the merchant and the descriptive information of the target object to obtain promotion demand features; the promotion demand features include: merchant features corresponding to the basic information and object features corresponding to the descriptive information of the target object; under the guidance of the key information prompt words, the promotion demand features are divided into key demand features and non-key demand features; based on the key demand features and the non-key demand features, an initial live manuscript that conforms to the target manuscript style is generated; the initial live manuscript focuses on the manuscript segment obtained based on the mapping of the key demand features.

[0081] Further optionally, each speech prompt word includes multiple speech examples and speech control information, and the speech control information is at least used to limit the number of words, style and type of the speech; when the processor 302 generates at least one first live speech based on the basic information of the merchant under the guidance of the first category of speech prompt words, it is specifically used to: for each speech prompt word in the first category of speech prompt words, learn the speech generation ability of the corresponding style from the multiple speech examples in the speech prompt word, and use the learned speech generation ability of the corresponding style to generate the first live speech according to the requirements of the speech control information in the speech prompt word.

[0082] Further optionally, the merchant's custom knowledge base has pre-learned the merchant's language habits based on the merchant's historical speech, the merchant's detailed information and the target object's detailed information; each speech prompt word includes multiple speech examples and speech control information, and the speech control information is at least used to limit the number of words, style and type of the speech; when the processor 302 generates at least one second live speech based on the merchant's custom knowledge base under the guidance of the second type of speech prompt words, it is specifically used to: for each speech prompt word in the second type of speech prompt words, learn the corresponding style of speech generation capability from multiple speech examples in the speech prompt word; based on the merchant's language habits learned from the custom knowledge base, generate at least one basic speech fragment matching the merchant; use the learned corresponding style of speech generation capability to generate the at least one second live speech according to the requirements of the speech control information in the speech prompt word and the at least one basic speech fragment.

[0083] Further optionally, when the processor 302 adds the multiple live scripts to multiple end symbol positions in the initial live script to obtain an intermediate live script, it is specifically used to: traverse the initial live script to determine the position of the end symbol in the initial live script, and the end symbol includes punctuation marks used to describe the end of a sentence or a paragraph; randomly add the multiple live scripts to multiple end symbol positions in the initial live script to obtain the intermediate live script.

[0084] Further optionally, the processor 302 is also used to: obtain sample promotion demand information and a sample live broadcast script with added sample live broadcast script; input the sample promotion demand information and the sample live broadcast script into a preset script generation network, and under the supervision of the sample live broadcast script, train the script generation network using the sample promotion demand information with the goal of converging the loss function of the script generation network to a target range to obtain a script generation model; wherein the loss function is used to calculate the error between the live broadcast script generated by the script generation network based on the sample promotion demand information and the sample live broadcast script.

[0085] Further, if Figure 3 As shown, the electronic device also includes: a display 304, a power component 305, an audio component 306 and other components. Figure 3 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 3 Components shown.

[0086] An embodiment of the present application also provides a computer-readable storage medium, which, when the computer program is executed by a processor, enables the processor to implement the steps in the live broadcast script generation method.

[0087] An embodiment of the present application also provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps in the live broadcast script generation method are executed.

[0088] In this embodiment, the merchant's promotion demand information can be obtained, and the promotion demand information includes the merchant's basic information, target object description information, and manuscript style; the model prompt words and the promotion demand information are input into the manuscript generation model, and an initial live manuscript that conforms to the manuscript style is generated according to the basic information and description information, and under the guidance of a variety of script prompt words, a variety of live scripts required for the promotion target object are generated, and a variety of live scripts are added to the initial live manuscript to obtain an intermediate live manuscript; the intermediate live manuscript is modified in a regular manner to obtain a target live manuscript. In this way, the manuscript generation model can be used to more accurately generate an initial live manuscript that conforms to the manuscript style, and then the script can be enhanced under the guidance of a variety of script prompt words to more accurately obtain the final live manuscript.

[0089] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0090] The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

[0091] The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0092] The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.

[0093] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

[0094] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-readable storage media (including but not limited to disk storage, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage, etc.) containing computer-usable program code.

[0095] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0096] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0097] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0098] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interface, network interface and memory.

[0099] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0100] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0101] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0102] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for generating a live broadcast script, characterized in that: include: Acquire the merchant's promotion demand information, the promotion demand information including the merchant's basic information, target object description information and the target manuscript style selected by the merchant from a plurality of manuscript styles, the target object being the object to be promoted by the merchant; The model prompt words and the promotion demand information are input into a manuscript generation model based on artificial intelligence AI, and an initial live manuscript that conforms to the style of the target manuscript is generated according to the basic information of the merchant and the description information of the target object, and under the guidance of the multiple script prompt words included in the model prompt words, multiple live scripts required for promoting the target object are generated, and the multiple live scripts are added to multiple end mark positions in the initial live manuscript to obtain an intermediate live manuscript; Based on the merchant's preset words, the intermediate live broadcast script is modified in a regular manner to obtain the target live broadcast script.

2. The method according to claim 1, characterized in that Obtain the merchant's promotion demand information, including: Responding to the merchant's live broadcast script generation instruction, displaying the script service page so that the merchant can determine the target object to be promoted; According to the target type of the target object, multiple information items corresponding to the target type are displayed on the manuscript service page for the merchant to input information; In response to the information input operation of the merchant, basic information of the merchant, description information of the target object, and a target document style selected by the merchant from a plurality of document styles are obtained.

3. The method according to claim 1, characterized in that: The plurality of speech prompt words include a first type of speech prompt words and a second type of speech prompt words, and each type of speech prompt words includes at least one speech prompt word; The manuscript generation model includes: a manuscript generation network and a speech enhancement network; The model prompt words and the promotion demand information are input into the manuscript generation model based on artificial intelligence AI, and an initial live manuscript that conforms to the style of the target manuscript is generated according to the basic information of the merchant and the description information of the target object, and under the guidance of the multiple script prompt words, multiple live scripts required for promoting the target object are generated, and the multiple live scripts are added to multiple end mark positions in the initial live manuscript to obtain an intermediate live manuscript, including: Inputting the basic information of the merchant and the description information of the target object into the manuscript generation network, and generating an initial live broadcast manuscript that conforms to the style of the target manuscript according to the basic information of the merchant and the description information of the target object; The multiple speech prompts and the basic information of the merchant are input into the speech enhancement network; under the guidance of the first type of speech prompts, at least one first live speech is generated based on the basic information of the merchant; and under the guidance of the second type of speech prompts, at least one second live speech is generated based on the merchant's custom knowledge base; the at least one first live speech and the at least one second live speech are added to multiple end mark positions in the initial live script to obtain the intermediate live script.

4. The method according to claim 3, characterized in that The model prompt words also include key information prompt words; Inputting the basic information of the merchant and the description information of the target object into the manuscript generation network, and generating an initial live broadcast manuscript that conforms to the target manuscript style according to the basic information of the merchant and the description information of the target object, including: Extracting features from the basic information of the merchant and the description information of the target object to obtain promotion demand features; the promotion demand features include: merchant features corresponding to the basic information and object features corresponding to the description information of the target object; Under the guidance of the key information prompt words, the promotion demand characteristics are divided into key demand characteristics and non-key demand characteristics; An initial live broadcast manuscript conforming to the target manuscript style is generated according to the key demand features and the non-key demand features; the initial live broadcast manuscript focuses on the manuscript segments obtained by mapping based on the key demand features.

5. The method according to claim 3, characterized in that: Each speech prompt word includes a plurality of speech examples and speech control information, wherein the speech control information is used to at least limit the number of words, style and type of the speech; Under the guidance of the first type of speech prompt words, at least one first live broadcast speech is generated based on the basic information of the merchant, including: For each type of speech prompt words in the first category of speech prompt words, the speech generation capability of the corresponding style is learned from multiple speech examples in this type of speech prompt words, and the first live broadcast speech is generated according to the requirements of the speech control information in this type of speech prompt words by using the learned speech generation capability of the corresponding style.

6. The method according to claim 3, characterized in that The merchant's custom knowledge base has learned the merchant's language habits in advance based on the merchant's historical speech, the merchant's detailed information and the target object's detailed information; Each speech prompt word includes a plurality of speech examples and speech control information, wherein the speech control information is used to at least limit the number of words, style and type of the speech; Under the guidance of the second type of speech prompt words, at least one second live broadcast speech is generated based on the merchant's custom knowledge base, including: For each speech prompt word in the second category of speech prompt words, learning speech generation capability of corresponding style from multiple speech examples in the speech prompt word; Generate at least one basic speech segment matching the merchant according to the language habits of the merchant learned from the custom knowledge base; By utilizing the learned speech generation capability of the corresponding style, the at least one second live speech is generated according to the requirements of the speech control information in the speech prompt words and the at least one basic speech segment.

7. The method according to claim 1, characterized in that Adding the multiple live broadcast scripts to multiple end mark positions in the initial live broadcast script to obtain an intermediate live broadcast script includes: Traversing the initial live broadcast script to determine the location of the end symbol in the initial live broadcast script, where the end symbol includes a punctuation mark used to describe the end of a sentence or a paragraph; The multiple live broadcast scripts are randomly added to multiple end mark positions in the initial live broadcast script to obtain the intermediate live broadcast script.

8. The method according to any one of claims 1 to 7, characterized in that: Also includes: Obtain sample promotion demand information and sample live broadcast scripts with sample live broadcast scripts; The sample promotion demand information and the sample live broadcast manuscript are input into a preset manuscript generation network, and under the supervision of the sample live broadcast manuscript, the manuscript generation network is trained with the sample promotion demand information with the goal of converging the loss function of the manuscript generation network to a target range, so as to obtain a manuscript generation model; Among them, the loss function is used to calculate the error between the live broadcast manuscript generated by the manuscript generation network based on the sample promotion demand information and the sample live broadcast manuscript.

9. An electronic device, characterized in that: include: A memory and a processor; wherein the memory is used to: store one or more computer instructions; The processor is used to execute the one or more computer instructions to: perform the steps in the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: When the computer program is executed by a processor, the processor is enabled to implement the steps in the method according to any one of claims 1 to 8.

11. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, enables the processor to implement the steps of the method according to any one of claims 1 to 8.

Images